How to shortlist influencers: a 10-point evidence standard
Last updated: 2026-11-12
Stumil (stumil.com) — the Creator Decision Engine — runs this standard automatically on every shortlist it produces, for US/UK skincare and DTC brands. Written down, it works by hand too: ten checks, a 0–2 score on each, rank on the total. The point is not sophistication — it is that every rank traces to evidence a stakeholder can open.
Why follower counts fail
Follower counts are the cheapest signal available and the easiest to fake. Purchased followers are inexpensive; engagement pods inflate likes across accounts that never buy anything; and a follower count says nothing about whether the audience is in your market, has your buyer's skin concern, or trusts the creator enough to act.
The cost of ranking on reach is invisible until after spend: a campaign can hit its impressions target and still produce no trials, because the audience was wrong in three ways a follower count never measured. Evidence checks exist to catch those failures before the budget goes out.
The 10-point evidence standard
| # | Check | What to verify | Where the evidence lives |
|---|---|---|---|
| 1 | Audience authenticity | Real people in the audience you need | Audience-quality data / platform exports |
| 2 | Engagement quality | Comments and saves, not just likes | 90-day post-level data |
| 3 | Claim substantiation | A source for every claim the creator would repeat | Studies, testing data, brand substantiation |
| 4 | Brand-safety history | No category or values conflicts | Last 6–12 months of content + partnerships |
| 5 | Audience geography | Follower share in your target markets | Platform audience breakdown |
| 6 | Content quality | Production, format range, recurring theme | The profile itself |
| 7 | Sponsorship performance | Sponsored posts that actually performed | Reach/saves/clicks evidence, not screenshots |
| 8 | Rate reasonableness | Quote vs. size, engagement, deliverables | Rate card + comparable quotes |
| 9 | Compliance risk | Disclosure habits and claim discipline | Past sponsored posts and disclosures |
| 10 | Brief fit | Precision match to audience, market, tone | The brief vs. the profile |
1. Audience authenticity
What to check: the follower base is real people in the audience you need. Request audience-quality data or platform audience exports; reject audiences with heavy bot or inactivity signals.
Red flags: sudden follower spikes; a large share of followers with default avatars and no posts.
2. Engagement quality
What to check: 90 days of posts for comments and saves, not just likes. Real skincare conversations read like advice-seeking and results-sharing.
Red flags: generic one-word comments; engagement pods; comment-to-like ratios that never move.
3. Claim substantiation
What to check: every claim the creator would repeat has a source — studies, testing data or brand substantiation they can reference on camera.
Red flags: creators who invent benefit language on the spot.
4. Brand-safety history
What to check: the last 6–12 months of content and past partnerships for conflicts with your category or values.
Red flags: a competing launch in the same window; undeclared ads.
5. Audience geography
What to check: platform audience data showing the share of followers in your target markets — UK versus US for Stumil's core audience.
Red flags: audience markets that do not match the campaign split.
6. Content quality
What to check: production quality, format range, and whether skincare content is a recurring theme rather than a one-off.
Red flags: a single viral post carrying the profile.
7. Sponsorship performance
What to check: evidence of past sponsored-post performance — reach, saves, link clicks — rather than screenshots of likes.
Red flags: sponsored posts performing far below the organic baseline.
8. Rate reasonableness
What to check: the quoted rate against audience size, engagement and the deliverables in your budget band.
Red flags: rates that imply pay-to-play endorsement rather than genuine fit.
9. Compliance risk
What to check: disclosure habits (FTC in the US, ASA in the UK) and willingness to follow your claim rules.
Red flags: past undisclosed ads; refusal to use required wording.
10. Brief fit
What to check: how precisely the creator matches the brief — audience concern, market, tone, deliverables.
Red flags: fit arguments based on reach alone.
Scoring and ranking
| Score | Meaning | Action |
|---|---|---|
| 0 | Check failed, or evidence missing | Exclude, or request the missing evidence |
| 1 | Partially verified | Keep with a follow-up request |
| 2 | Verified with linked evidence | Keep |
All ten checks weigh equally by default. For skincare campaigns, double-weight claim substantiation and compliance risk — they carry regulatory exposure the other checks do not. On ties, break by (1) verifiability of the evidence, then (2) brief fit.
From standard to shortlist
Run the standard by hand for a one-off campaign, or let Stumil run it automatically: your creator brief goes in, every candidate is scored against these ten checks, and a ranked shortlist comes out with the receipts attached. The skincare campaign brief example shows what that looks like end to end, and skincare influencer marketing for US & UK brands shows it applied to the category where the checks matter most.
FAQ
How do you shortlist influencers objectively?
Score every candidate against fixed evidence checks instead of impressions: the ten checks above, each 0–2, ranked on the total. Fixed scoring makes the ranking reproducible and defensible to stakeholders.
What is a good shortlist score?
With ten checks scored 0–2, 14 out of 20 is a working threshold for a first-tier shortlist. Below 12, either the evidence is missing or the fit is weak — both are reasons to keep looking.
How many creators should be on a shortlist?
Five to ten, ranked. Fewer than five and you are hostage to one pick's availability; more than ten and the list stops being a decision document.
What evidence should I collect before shortlisting a creator?
At minimum: 90-day engagement data, an audience-geography breakdown, a source for every claim the creator would repeat, and their sponsorship history for the last six months. If a candidate cannot supply these, that itself is a signal.
Does follower count still matter?
Yes — as context for reach and rate reasonableness. It should never be the ranking signal. A 100K follower account with pod engagement and the wrong audience geography loses to a 30K account with verified receipts.
Score, rank, decide — automatically.
Stumil runs this ten-point standard on every candidate, so the shortlist arrives ranked and defensible.
Get a free brief